Phison Electronics Corp. (8299) Earnings Call Transcript & Summary

Hello, everyone. Welcome to Phison's Third Quarter Earnings Call. This is Jason Tsang from CLSA. Today, this is our honor to have K.S. Pua, the CEO at Phison here with us. K.S. will share current update on outlook of both NAND flash industry market and Phison itself. After that, we will open the floor for Q&A section. Now let me turn this call over to Khein-Seng Pua, please go ahead.

Yes. Thanks, Jason. Thanks, everyone, to join our Phison's CQ3's earnings call. Okay. Since this quarter, we host only one section. So in some portion, I will give both language, English and Chinese to make sure not any misunderstanding of my statement. Okay. This is our CQ3 is our revenue breakdown. In this chart, you can see consumer still, to be honest, CQ3 and CQ4 consumer is still weak in demand. You can look for the whole industry, PC, IoT, smartphone is kind of weak. But at least this business-wise, we keep our best to keep our share. But in controller-wise, we still keep flat on our controller revenue. And CQ4, we can see some PC OEM increase their portion in the controllers. Industrial-wise, since you can look industry actually weak for the other 4 quarters. Recently, industry demand is slightly back. Gaming. Gaming is really ugly, like refer to the consumer inflection. So gaming, we almost dropped only like 1/4 of the 4 quarters ago revenue. And enterprise, since we invest a lot, we're cooking a lot, enterprise is around 16% of revenue. Next year, we were going to have a big growth on our enterprise business. That I will break down to you. And again, if you look to the whole industry, the consumer-wise, especially in the SSD for retail is really demand, is very weak. And I believe I don't see the big demand will be back because if you buy a laptop, a desktop with the SSD 512 gigabyte 1 terabyte you are not going to buy the 5 gigabyte from retail. It's going to the 1 terabyte, 2 terabyte. So the unit-wise will be declining. And they will ask mainly is high-quality, high performance and high brand, the good brand products. So what is the trend for futures? Futures, if any NAND company, you go to the NAND supplier, everyone going to enterprise. In the SSD, everyone going to the AI applications. And Phison, we are direct go to the PC OEM and also to the mobile brand company. Without that, you cannot move any of your components. And Phison has the best advantage is in the customized storage solution, mainly for automotive, for industry, for security, for auto space, the business. So Phison since 3 years ago, to be honest, we realized when the PC OEMs start to put 5t into the PC, this is coming. So we prepared a lot of new solution for our future, okay? So our CQ3 result revenue-wise is slightly up Y-o-Y, but declining Q-o-Q because of very weak in the consumer market. But of course, we gained a little bit in the enterprise share, but still not enough. By profit-wise, Y-o-Y 2.2% up, but Q-o-Q, we're down 26.4%. Reason is the revenue dropped a lot versus CQ2. So revenue-wise, you can see from the NTD 16 billion down to NTD 15 billion, now down to NTD 13 billion. The reason is we don't want to go to follow the price war. We insist to follow our position to get more business with the customers with the design project. Gross margin-wise, okay, overall CQ3 29.2%, but we try to count back what happened to this 29%. We have our write-down on inventory, which impact gross margin 3.3%. So if deduct this, the operating margin is 32.5%, which is a little bit slightly lower than the CQ2. CQ2 is around 34%, so around 2% lower. But since from the market, you know the NAND supply, they are now starting to ask low aggressive price. We keep buying the inventory. But when we're buying that, we need to write up our inventory. So this is the reason why we are below 30% since these 8 quarters. [Foreign Language] Earning-wise, the CQ4 3.37, I need to break down what happened here. Write-down impact close to the EPS NTD 2 means it's around NTD 400 million, okay? And loss on finance means is our investment to the public market, which is minus because CQ3, the stock market-wise is kind of weak in the memory business. We have to invest in peso and the OSC. We have to cover the loss for NTD 1 EPS. Foreign exchange, especially in the September, the NTD raised a lot is our loss is close to the NTD 1. So if we add back this is around close to NTD 4, we are around like NTD 7- something EPS, which is close to our CQ2. CQ2, the reason we have some investment gain. So overall, by operating profit, CQ3 declining because of the write-off and also because of the revenue. [Foreign Language] Our income statement. Revenue-wise slow because the market is slow. And overall, when we look to the operating profit, operating profit-wise, by CQ2, we are declining just because of the revenue, the impact by revenue. Look to our expense. Actually, by CQ3, we are controlling our expense, especially in the R&D expense. So R&D expense is the lowest in these 4 quarters. okay? I think the most impact is coming from the nonoperating gain, mainly inventory mainly the write-down and also the investment loss, okay? And overall, by the OP margin is around 8% versus the CQ2 12%, but better than last year CQ3. But again, overall, we are still working hard to try to make sure this year, we're able to get the much better profit to contribute back to the shareholders. Okay. Back to the balance sheet. Inventory-wise, we are slightly declining the inventory not because we are buying less. We keep buying more quantity, but the price from supplier are declining; means the inventory by pay actually increasing, but the unit cost declining. Then the question is why we are keep buying? The reason is enterprise business, we need to accumulate our inventory for 2025 year. And NAND vendor now more willing to offer lower price to us. [Foreign Language] I actually is not much different, okay? So it's slightly different because of some amortized of employee stock option, but this is a small number, okay? So I think I can skip this one. Okay. October revenue is $3.7 billion. I think you may surprise. The reason is the demand this year, we don't see the Double 11 demand. We don't see the Christmas Black Friday demand. And this NTD 3.7 billion, most likely still coming from the industrial business, enterprise business and also the, mainly is embedded ODM business. We are not going to follow the consumer market because they want us to. And if you can read some report from the China module house, they are turned to loss since CQ3. And I believe CQ4, they will turn to much bigger loss, okay? [Foreign Language] In auto wise, this year, we work very hard. We get extra 20-plus of a new design in mainly from the U.S., Europe and the China market. And this business only you get the design when it is 5 to 7 years longevity supply. So the price is fixed. So this one can help us to bring much better gross margin in the future. So we also get some new products in our U.S. 2024 in the automotive. And also, we are engaging with our NAND semicon to get more project and design win together in automotive. So this event just happened in the Detroit. This is automotive, we invited by the association to give a speech and to tell them what we did in the storage automotive, okay? This is to share with you. Also, Phison, we are proud to get the world's first NAND controller independent supplier to achieve ISO/ SAE 21434, the certificate, okay? We are the first one to get the approval, and this also can help us to gain more business opportunity and design win in the coming future. Okay. Everybody may agree in the semicon today, only AI business is doing good. Other than AI, consumer, medical, everything weak, okay? But the question is, since AI invests a lot we can see a lot of new data center under construction. This is the source we get from the market. By 2024, they still have around 200 plus of data center under construction. Why? Because of the AI application. When owns the data center ready, they are going to buy the equipment, the GPU, the CPU and the storage. So this is what happened here. In the last 2 years, last 18 months, the CSP spent over $100 billion to buy the GPU. They need the GPU come back as an ROI, return of investment, they need the users. So users, once the users pay and use the AI, then we generate a lot of data, then they need data center with the enterprise SSD. [Foreign Language] The new business of Phison AI is adopted. We are promoting very hard. We have a big investment. We have many product line in the business. Recently, we also got a lot of inquiry. We are not only selling the component, but now we are providing the server with our software application. But now we are facing 2 difficulty, okay? The one difficulty is we expect the software company since early this year, they come to our ecosystem. They can build their software to support this ecosystem. But unfortunately, the market don't have enough of AI engineers, they also don't know what kind of business they can approach. So Phison, we have to develop all the software by ourselves. [Foreign Language] [Presentation] Phison Electronics aiDAPTIV+ hybrid solution now launches a powerful feature real-time personal AI assistant available on instant messaging platforms. The AI assistant can receive commands and utilize a pretrained model to generate presentation files that meet specified requirements. It can also customize the presentation layout according to the unique design needs of different industries. The AI assistant can automatically compose e-mails, generating content that is tailored to specific needs and contexts. It streamlines communication by drafting e-mails efficiently, ensuring that the tone, style and information are appropriate for the intended recipient and purpose. The AI assistant can automatically analyze YouTube video content, identifying key themes, extracting valuable insights and summarizing essential information. Using advanced natural language and video processing, it efficiently breaks down complex content into concise, understandable formats, making information retrieval easier and faster for users in various industries. The AI assistant can read and analyze documents automatically, extracting key information, identifying patterns and summarizing content for quick insights. Leveraging natural language processing, it transforms complex documents into easily digestible formats, enhancing efficiency and supporting informed decision-making across various industries. So recently, we got around 30 set of POC order from our corporate customers, which is a server, including with Phison, our software. Recently, these few new features we are going to promote to the market to help the market more easier to adopt the generative AI on edge solutions. We also start to work very close not only in Taiwan, but also in U.S., Southeast Asia and the Europe customers. We also success to attract the GPU makers to have a POC with Phison. And this is since they have NDA, we are ongoing. Hopefully, by end of CQ4, we are able to launch the products together with them. This is all the software, all the features developed by us. Phison, we also took 3 hospital projects in Taiwan, helping the hospital to build their on-prem AI solution. We're also working with our Miaoli Polish station to help them to install one AI equipment here to help them to doing, assist them to doing the document. And this is the, a lot of features we request by the corporate customers. So Phison now, we try to promote aiDaptive. In the past, we hope the software company can help to develop software. But unfortunately, Phison had to build everything by ourselves, and we did by ourselves. [Foreign Language] Second challenge, we have a lot of corporate business in worldwide. They like to use our solution. They like to buy the server, the software from us, but they tell us they don't have engineers, capable engine to handle. So they ask Phison to send our engineer to their office or we need to train the engineers. So we realized this market, we don't have enough of AI talent. So we developed the so-called AI training PC, by CES, coming CES, we can, I mean, the market can see the GPU company is going to promote this AI training PC in the market. Why? This is mainly for education. Imagine one thing, every university, they want to have AI class. Every student, they like to have a GPU machine to learn the AI. But unfortunately, GPU machine too expensive. And this machine, we aim to USD 4,000. So school can buy. The parents also can buy for their kids. So we already got the, attract the GPU makers to co-work on this platform. And to move to the volume-wise, this is much easier and much faster. So we hope, we Phison we donate this to our Taiwan Yang Ming Chiao Tung University. We're also going to talk to the NTU and CKU and THU. We believe this can go popular soon to help to train the talent immediately from the university and also can help to train the talent in the market. [Foreign Language] So back to our, the current OpEx, actually, by CQ3, we are controlling our R&D expense. So overall, the expense is lower than our last 2 quarters. But anyway, we believe by CQ4, we still can be able to earn good enough profit to us. So we're still able to have a good revenue and good profit and good cash dividend to our shareholders. Okay. Let me have some summarize. Overall, the consumer market is very weak. The consumer retail SSD is super poor because of not necessary to buy from the market. With the AI solution, we get a lot of good traction from the market, not only the enterprise customer, but also the GPU makers. We engage with the one GPU makers. They send the server to us. We are ongoing, doing the POC. Hopefully, by end of December, we can launch together. And for the AI training PC, this is one can help us to gain the faster revenue because much easier to adopt by the market. So enterprise SSD, we get at least 5 box maker in U.S. We engage with them. We customize our solution to them. The revenue-wise will start to happen by end CQ1 starting by early CQ2. So this will be our revenue generator and our profit next year. Last, Phison, we start to build a team to develop our in-house software for aiDAPTIV+. I don't like to do that, but I got no choice because without the software, we can't able to sell more of our solutions. But with the software ready, the software is not going by free, it's by license. So the margin of software can be 100%. So next year, we expect when the AI training PC popular, more talent trend from the school and from the market, we are able to sell more our solution to get more hardware revenue profit and also the software profit. So this is summarized. So a few Q&A questions. Let me finish, then we can go to online. Inventory-wise, we start, we still accumulate the enterprise our inventory, especially in the controller and NAND. The NAND is for the very specific NAND partners, which they cannot supply you in 1 day. So we start to accumulate for our next year use. Industrial-wise, demand turned to strong. So we need to keep some -- mainly for automotive. There are some components going to EOL. So we need to accumulate 7 years inventory. Embedded ODM, we are now working directly with the PC Tier 1 OEM. We need to ship to them next year first half component. If price, NAND price is good enough, we need to start accumulate. And the retail, retail is very weak, but we still need to have maintained our share, USB memory card and SSD, and also the gaming one, okay? The news saying that regarding the production cut, according to our contact, the answer is yes. The cutting is happening. Of course, they have a rumor saying that the cutting is really impact the market or not. But let's see by December. We believe December, more NAND vendor will jump in to say they are going to cutting the production. The reason is very simple. In the last 5 quarters, they lost NTD 40 billion accumulated in the NAND business. In the CQ1, CQ2 this year, business is good, but they're not earn back NTD 40 billion yet. So I can expect they don't like to keep losing money. And by the way, next year, the enterprise for AI is strong. They also don't like to see the price going to minus margin. So cutting is going to happen. So this question is when is consumer market coming back? For retail SSD, I think it's gone. Gone doesn't mean it's nothing. Gone means it's going to be very niche because it's not necessary to upgrade 100% or maybe only like the PC users, 5% upgrade their SSD from 512 to 2 terabyte, means they need the quality, they need the brand, they need the service. And Phison, we are good in this business. We are not the price killer. Inflection, actually, if you all agree, other than AI, is very, very weak in the business. So that's why we are working hard to get the aiDAPTIV+ business and also enterprise business. Next year, U.S. election is done. So we don't know what's going to happen, either inflation up or inflation down. But no matter what, Phison, we always make ourselves ready for our next challenge. Good thing is aiDAPTIV+ getting much smooth, highlighted by the CPU company, we have a POC. Enterprise SSD, we got lots of projects ongoing, which can contribute the revenue. Customer module, they have a lot of customer module ongoing by [indiscernible]. And last is we are moving to the software business. Hopefully, software can contribute some 100% gross margin to us in next year. Okay? So this is my presentation.

[Operator Instructions] So now we start the Q&A section. So my first question is, can we provide the outlook or guidance for Q4 regarding the each of the sectors, including consumer, industrial, gaming or embedded ODM?

Okay. We have a system. Every month, we have a system to show me the order we collect. In this year, the lowest order collect was happened by September, okay? October, the order we collect by dollar amount improved 20% versus September. November, we still try to collect more revenue. Of course, when I say that no Black Friday, but still they got some last order from gaming, from PC, from some high-end products. But in enterprise-wise, we're still gaining our business. But enterprise-wise, recently, we aim to mainly AI application, which is a high-density products. This ongoing project with the customers, this revenue will happen by end CQ1 next year. So overall, by CQ4, controller, I think demand is okay because the PC -- our NAND makers start to pull in controllers. Industrial-wise, it's good. Gaming is a question. I don't know if Christmas really improve the gaming demand. OEM is okay. But overall, CQ4, I believe the goal is to keep flat of CQ3 revenue. This is the goal, okay? Unless we are willing to cut price to get an order unless -- but not much wish to cut price. The reason is we're still collecting the inventory for our enterprise and embedded OEM. So overall, the wish is to keep flat by CQ4. Okay.

Okay. My second question is in terms of your gross margin. In Q3, you mentioned that around 2% gross margin was impacted by inventory write-off. So which means another around 2% gross margin is impacted by probably your product mix or your pure product or ASP. So my question is that you mentioned that your pricing in Q3 or even in Q4 can still maintain in a stable level because you don't want to join the price war. And your product mix in Q3 seems better because your industrial and enterprise remain good. So I just wonder which factor impact your gross margin by around 2% in Q3? And how do we look at your gross margin in Q4? Without inventory write-off, yes.

Yes. Q4, the goal is flat as Q3, the goal. The reason is, first, we have some high-cost inventory happened by Q1, Q2, but write off something in the CQ3. You can see around 3.3%, write-off. So some high-cost inventory start to write off, and we are buying the lowest price from the market. One thing is the market is low, right? We, selling price definitely have to go low. But I don't want to go that crazy low to minus. When I look to the module house in China, since last, since May, June, they start to be very aggressive in price, cutting price. Now CQ3, they lose money. CQ4, I can tell you, is a bit loss on their business. But Phison, luckily, we have some high mix, industrial enterprise. The goal is CQ4, we like to keep flat, okay? We also can expect some NRE coming from our design service customers, okay? I believe CQ4, since the market is slow, we think the NAND manufacturer may keep telling the market they are going to cut the production. If this is true, then we need to see the demand will turn to good or the confidence will turn to good by next year CQ1. Okay.

We had one question from our investor. The first one is, if our AI SSD, I mean, the penetration rate can be largely adopted by market, will other competitors or NAND manufacturer release those kind of similar products in the future?

Since this is in the memory, I use memory answer, okay. [Foreign Language] So I just explained aiDAPTIV+. I don't think the NAND company, they're willing to do by themselves since Phison announced this is our product. And we have already got one NAND company come to us to work to ask how Phison to help them develop this, they just consign flash, okay? So we believe if NAND company willing to do this, we can take the business. China, we believe a lot of module house doing the same thing, but this is not easier because they had a lot of software activity. We have a Taiwan partner work with us taking our product. Now they try to develop by themselves. We give a message to them is simple. Once this happen, we will take any action to protect ourselves if they want to go to that way. But I don't think they are willing to go to that way because we will strongly protect our benefit.

Okay. And last question -- my third question is in terms of your visibility on the NAND price trend. I mean, currently, we do see NAND spot price or module price is declining. And you also mentioned that you expect that probably NAND will shortage again in second half next year. So when do we expect that NAND price can have reversions to expect meaningful growth on no matter spot price or contract price in the future?

Okay. This market is very transparent. You go to see what is the market price by CQ3 and you defer to the NAND companies' financial book, you can easy to know the gross margin, okay? And by every month, the price declining, maybe 2 to 3 months, the gross margin will turn to minus. I mean, component in wafer wise. So I can see the NAND company, that is not willing to cut the price under margin. On the other hand, they're selling the enterprise SSD they still enjoy very good margin. Why they keep selling the wafer to the market? Just because the consumer market is so slow, is so slow. In this case, right, in order to protect their benefit, their profit, they may start to reduce the shipment of wafer to the market. They have to make sure they drive enterprise, drive market, the price can keep stable. So we also can see the enterprise users, they have a demand, but they push out their purchase order to next year. They also expect the NAND company willing to cut out the price. But if by December, if NAND companies start to suffer the component minus margin, they start to announce to cut the production, then the buyer of enterprise, they may to think, do they need to buy now or they push out? If this is the case, right? This may create a kind of phenomenon, the demand turn to strong on enterprise, okay? So I have no comment yet, but we can see by December, the NAND company, what they are going to do. But I have a strong confidence, cut production is going to happen if the price keep declining.

The next question is in terms of your enterprise. It seems like you get so many design wins and market shares currently. So how do we expect that your enterprise contribution in the future? And we know that enterprise gross margins or profitability is significantly better than your consumer products. So how do we expect that it can probably boost the gross margins or profitability in the future if their contribution is getting bigger?

Okay. Phison's Enterprise SSD, we have a 2 business model. One, we sell the drive. Second, we're doing design service. We're helping our customers doing the design, the whole drive with our certain agreed gross margin. okay? So when we talk about enterprise, in the whole enterprise, the drive is maybe let's say the 64 terabyte drive, the NAND component price is around 90% of the drive. So this business will be kind of a 3-party business. We work with our customers. Customers, we both work together with the NAND company. They negotiate the price. We take the waiver, take the flash, helping them to build the drive with our certain agreed gross margin, okay? So I want to just clarify, Enterprise SSD don't expect the gross margin can be 40%, 50% impossible because the NAND component is 60% to 90% of the drive. But we are able to improve our buying power. There, we can use the same value of component price to sell to others -- the Tier 2, Tier 3, Tier 4 with the customized design, we can get our extra profit. So I'm hard to answer you what kind of a gross margin is going to happen on enterprise, but this can create first, our revenue, our net profit and improve our buying power. And by the way, sorry to say that, if we are in the NAND business, we cannot rely on USB, SD, virtual SSD anymore because it is gone. Since we need to grow ourselves, Enterprise SSD is a must to get into this industry, okay? I believe with enterprise next year, we grow revenue, we grow net profit.

My next question is also in terms of your data centers or automotive opportunity. I think personally you released your re-timer, and I think you also have a very good design capability on your controller. So how do we expect that your data center business can probably bundle those -- your good quality products on the logic products with your modules or memory products in several data centers, also auto as well?

Okay. Again, the whole market other than AI, everything is slow. Even today -- I mean, since the last few quarters, data centers for storage are also slow because all the budget of a CSP, they go to buy GPU. But sooner or later, they start, now they are focused on storage because without storage they cannot generate their revenue and their profit. Okay. So look to this industry, re-timer, we are late comers but we gained a few projects. We work with a CPU company, we work together with them, we also work with their server motherboard makers. We get a few projects from them, and this gross margin is much better than any other business because this is niche and we have a good quality of products. Second, for data center, the SSD, we're not able to get the Tier 1 data center because we are small. So we work with our U.S. partner who is making the box. We design the enterprise SSD for them, we build a drive for them. Then they sell to the data center with our drive -- with our service. But we -- now Phison's policy since this CQ3, we start to engage with the end users, why the end users? Who buys the equipment, installs it to the data center, we directly work with them. This is just because we want to shorten the design process, the relationship, and increase our inference power. We need to let them know who we are, they're happy with us, then they prefer to use their assembly house to use a Phison component, not only re-timer, but also the driver, our SSD, our drive. So this business, we need time to invest, we need time to cook, but we're already there. Again, let me repeat. If we are in the NAND business without going to the data center, we are going to the AI, to customers, gone, finished. Because you can't see the future in retail.

Okay. We got one question from Simon.

Can you hear me well?

Sorry, I'm now using the PC.

Great. So Mr. Pua, would you recap why the inventory write-down happened? Because I raised the same question a year ago when the NAND price was very weak, but you said no inventory write-down. But today, still the NAND price is okay, but why do we have to see the 3% of the revenue is off?

I don't know where you see the NAND price, okay, but I can share with you the price September -- September versus June, the price I purchased, I got 45% discount. This I can share with you, I got a 45% discount. What you see today is not NAND price, you see enterprise SSD price. SSD price is good, but component price is very weak. Actually, my supplier told me by September, when they sell me those [ wafer ], they said they are in minus margin.

Okay. But back to the -- a year ago, you remember the -- like 2023 second quarter, third quarter, quarter-on-quarter basis, NAND price also collapsed a lot. And then the Q-o-Q also reported a record high amount of the loss at the time. How do you compare current pricing momentum?

In this, you see the mouse -- the cursor, right?

Yes, yes.

CQ1, CQ2 component price is high. The PLC is around [Audio Gap] per gigabyte, okay? And here, I'm telling you I got 45% discount. So from that one, I need to write off, okay? Back to last year. Last year, the worst high cost happened by '22, CQ3, CQ4. In '23, CQ1 already touched bottom, component price down bottom price up. So no write-off. Write-off only happens from high prices going to down prices like CQ1, CQ2 now. So it means by CQ1 next year, we don't need to write off. We may write back.

Okay. Yes, the follow-up question is when we look at the spot price, still the NAND price is getting weaker. So should we expect a further inventory write-down for December quarter then?

Okay. It depends on this, we call offset. On the book, we have close to NTD 2.3 billion reserve. When we start to use those component write-off, then the reserve will coming back to the book. And we buy something cheaper, then we may write off. Sorry to say, I can't tell you. But in the October, since we got some old inventory we sell to the market, we write back 5% of gross margin. So to be honest, I can't tell you what is going to happen because this little bit complicated.

Yes. It's okay. If I ask some very quick question, sorry. October revenue collapsed were much lower than the market expectation. But meanwhile, you are saying December quarter revenue likely flat quarter-on-quarter. That means you have to sell a lot of yes, here we go. So you have to sell a lot of your products through the November, December.

So do you see My goal. We are working very hard. But again, the market other than AI is very, very weak. You look to the Korean business, right? Hynix is doing good. The rest of Hynix other than HBM, other enterprise SSD gone, right? Bad. So I'm not complaining. This is all about high inflection, but we're still working hard on adaptive. We're still working hard on enterprise SSD, try to see if we can improve. But next year, next year, definitely will be good.

However, you mentioned that there are some software problem. So we have to wait to see the better software to promote your aiAdaptive+, right?

Yes, true, true. I also feel very painful because a lot of software company, they didn't pay their time to develop we try to do everything by ourselves. This is something like NVIDIA 15 years ago, they started the CUDA, right? But we start to build our fundamental software. On our software ready, we can gain extra gross margin, 100% gross margin.

Very quickly, Mr. Pua, AI adaptive revenue now is 0 or 1%.

No, no. Still there. We sell our AI100 SSD to the market. But again, it's very difficult because of software. So now we are selling the whole server. CQ3, we shipped over 30 server to our customers with our software and the margin is much better.

So this portion out of the total enterprise solution, I remember you have about what high-teen percentage of the total revenue is enterprise solution. What percent?

Yes, this is server wise is still small because only like 30-plus, right? But I want to clarify, this AI training PC can move much faster because this is only like $4,000, $5,000 products, okay? There will coming CES, the GPU company is going to promote this.

Okay. So what percentage of this revenue out of the total enterprise revenue?

Still small. Forget it, still small. It's not big. But we are going to disclose this next year when we start to see the volume coming cooking, we are still developing.

So 2 things. Last year, your company okay, profitable OP margin, high single digit. So even the NAND price currently weak, but we don't have to worry about the margins for the EPS number generate. Positive. No need to worry about negative margin, right?

No. CQ4, our operating margin is still positive. Actually, since October, we have a write-back. We sell the old inventory, which went off then the write-back we're doing good. We're doing good. So Q4, I'm still optimistic on our net profit because Phison still have other many investments is with a good return.

But the investors' concern is sort of the downward trend. Your gross margin was so great last year when the NAND makers really suffered. And then Q1, Q2 is so great. I remember you emphasized record high revenue, record high gross margin, but eventually, your margin now shaky, squeezing. So this is kind of the 1 quarter phenomenon or maybe at least 2 or 3 quarters of winter season or.

I think 2 quarters.

2 quarters, so which means.

We are not gut. We cannot match what's going to happen, right? If your Korea industry, Hynix enjoy others*, right? We work very hard to keep company profitable. By the way, we keep investing for future. So I am to say that this margin maybe like 2 quarters. But next year, when the new product is ready, everything good, we bad. But you look to the other competitors without the enterprise, without the AI, without customer product with automotive, how can they survive. If they cannot survive, they are suffer. That means we are able to get major share, right? Look to Hynix, 5 years, 7 years ago, Hynix, how can they compete Samsung, but you see what different? Yes, yes. More further, look more further, okay? We keep investing. When people love Hynix, they are saying something stronger. 5 years, look to the Intel AMD, right?

Yes. But we never seen like below the 30% gross margin for the past few years. And then if this sub 30% gross margin continues into December quarter, March quarter, it's kind of a little headache.

It is what it is. But at least I'm telling the shareholders, look at the future.

Okay. So you're going to return continuously 60%, 70% EPS for dividend, right?

Yes, yes, policy never changed.

All right. Sorry, I took too much time. I pass over to the operator for other guys question opportunity. Okay.

One question on the, [Foreign Language] So the question is asking about next year, is that Phison able to get the NAND companies' enterprise controller? The answer is yes. It's ongoing. Okay.

Yes, there's no more questions. So I have another question is that how do we look at OpEx for Q4 or in next year? Do we plan to hire more employees or headcounts in the future?

Okay, We have around 3,000 engineers in the Phison headquarter. We keep recruiting, but not heavy, not crazy recruit. We internally push the engineer team to use our adaptive to improve our efficiency. So by overall, the headcount increase is only single digit, okay? On the meantime, Phison also every month, we review the engineering team if the low performers will let go. So this whole year, the headcount increase by R&D is a single digit. Next year, it looks like most likely it's a single digit to 10% but follow what the market status Yes, last year, I will say '24, we are going to control our OpEx and we make it.

Got it. Got it. And my next question is in terms of geopolitical concern. Do we see any concern or risk due to the geopolitical or will U.S. government ask us to relocate our module production line out from Asia to U.S. Will that lead to a higher cost of goods sale to us?

Okay. Phison my plan. First, we have a China investment Hopefully, they can go public next year, then we can get our market share and also our return, okay? And China, use China, this is what their policy. We can't do anything, but we can use subsidiary to get the business there. India try to learn the China way. So we have a joint venture in India, helping our JV in India to get the local domestic business. And this is ongoing. We start to get adopted business from local enterprises from local. We believe this company can grow. And of course, the company, the goal is to go to public to become local company, Phison can get business share and also return of investment. And we also built our second I can call it headquarter in Malaysia because asked by U.S. customers. Taiwan, they say it's risky. Malaysia start to establish the R&D team. We have around 50 headcounts start to develop them, train them the enterprise SSD firmware development. So we are going to use Malaysia as our hub for future Southeast Asia business and Middle East business. U.S., Taiwan cover the U.S. business. And you're talking about manufacturing in U.S., it's simple because Phison we need only assembly, and we have a lot of U.S. local partner. If you're saying the cost is higher, we see higher price. But at the end, inflection is happening again, if this is a trend. But the good thing is Phison, we have a very special position. We are able to provide every storage solution to the market. So customers need Phison.

Got it. I have no more questions. And I think due to time interest, I think that's the end of our call today. Thank you, everyone, to attend our call today, and we'll see you next time. Bye-bye.

Okay. Thank you.